Seasonal ice dynamics control the timing of crevasse drainage at a fast-flowing outlet glacier

Crevasse field drainage transfers at least half of the seasonal runoff from the surface to the bed of the Greenland Ice Sheet, but the patterns of drainage are complex and spatio-temporally heterogenous. To better understand controls on crevasse drainage processes, we use an automated deep learning method to map the seasonal filling and drainage of water-filled crevasses at Sermeq Kujalleq (Store Glacier) in 2019 using 10-metre-resolution Sentinel-2 MSI imagery, alongside RACMO reanalysis melt estimates and strain rates inferred from ITS_LIVE velocity observations. The timing of crevasse drainage correlates strongly with seasonal peaks in surface strain rates, suggesting that seasonal variation in surface stress exerts a strong control on the exact timing of drainage to the bed. Linear elastic fracture mechanics (LEFM) modelling suggests that a seasonal transition to higher tensile stress regimes can trigger rapid full-depth hydrofracture at the point at which stresses exceed the threshold necessary to initiate initial fracture, provided that meltwater from early-season snowmelt is abundant. We suggest that this causal behaviour is distinct from both (i) supraglacial lake drainage, which depends on short-term (hourly-daily) transient accelerations to trigger drainage; and (ii) crevasse fields that are not observed to pond, which exist in high tensile regimes and likely readily supply water into the englacial or subglacial system from melt-season onset. The distinct mode of drainage described here requires unique parameterisation to predict spatio-temporal patterns of water transfer to the bed of ice sheets and drive models of subglacial hydrology.